Motion Models (Nived Chebrolu)

[@sid9424]

Great lecture with perfect explanations, much appreciated!

[@codebot2235]

Great lecture and very well explained

[@obensustam3574]

Great lecture

[@amuravindhiran1494]

Thank you for the lecture. Having a question on the material- why is the total probability for odometry model taken as product of individual probabilities when the noises associated with each of those individual motions are correlated?

[@teetanrobotics5363]

please make a separate playlist for these videos

[@RaushanKumar-bo2md]

Awesome explanation

[@lwufkudl]

Thanks for the lecture, I have two questions if I may :
1. Can we sample directly from the bicycle model (for the case of Ackerman drive) instead of using this abstraction of (rot1, trans1, rot2)?
2. In the two techniques presented (histogram representation and sample based implementation) how can you use such representations in the Kalman filter? The filter requires a Gaussian distribution with (Mu, Sigma), so do we have to approximate the banana shape distribution (both sample based and histogram) by a Gaussian?

[@sa-vx5vi]

for q1. wouldnt you still be doing same calculations but different sampling time? it is mentioned at 40:25

[@mohammadhaadiakhter2869]

Can someone please elaborate difference between hypothesis and odometery at 25:10

[@sid9424]

The odometry values represents the pose you would expect (i.e. without noise) after taking a certain action u, while the hypothesis is the corresponding Action including noise.